Monolayer Formation of Molybdenum Carbonyl on Cu(111) Revealed

Apr 30, 2012 - structure compatible with a (√7 × √7)R19 superlattice on the copper (111) plane. The arrangement and orientation of the molecules ...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/JPCC

Monolayer Formation of Molybdenum Carbonyl on Cu(111) Revealed by Scanning Tunneling Microscopy and Density Functional Theory Peter Krüger,† Mikhail Petukhov,*,† Bruno Domenichini,† András Berkó,‡ and Sylvie Bourgeois† †

ICB, UMR 6303 CNRS-University of Burgundy, F-21078 Dijon, France Reaction Kinetics Laboratory, INC, CRC of the HAS, University of Szeged, Szeged, Hungary



ABSTRACT: Molybdenum carbonyl Mo(CO)6 was adsorbed on the Cu(111) surface at 160 K in the monolayer coverage range and studied by scanning tunneling microscopy. A well-ordered monolayer of hexacarbonyl molecules was observed experimentally for the first time. The monolayer has a hexagonal structure compatible with a (√7 × √7)R19 superlattice on the copper (111) plane. The arrangement and orientation of the molecules on the surface were determined by density functional theory calculations, including van der Waals interactions. The comparison of adsorption and cohesive energies reveals that the molecule−substrate interaction is stronger than the intermolecular one, which explains the observed two-dimensional growth.

1. INTRODUCTION Metal carbonyls are a class of organometallic complexes that has a large application potential in nanotechnology. The interest is based on chemical and physical properties of the dmetal compounds: the particular arrangement of π and σ orbitals of the metal carbonyl molecule does not lead to a charge redistribution between the metal and carbon monoxide ligands, and the metal atom stays neutral if one or more of the carbon monoxide groups are lost.1 This nature of the metal− ligand bond makes the carbonyl molecule unstable under various physical processes, such as thermal activation,2−4 light irradiation,5−7 or bombardment with charged particles,8,9 whereby the molecule's decomposition results in clusters of metal atoms and desorbing carbon monoxide. The fact that the majority of metal carbonyls sublimate at relatively low temperature and can decompose under photon and electron beam irradiation makes these compounds attractive for thinfilm deposition and nanofabrication. One of the potential applications is the lithography by photoirradiation or electron beam exposition in vacuum conditions, where the desired pattern of metal clusters at the micro- and nanoscale can be created by controlled decomposition of metal carbonyl vapor above a substrate.10,11 Such a technique of electron beam decomposition of W(CO)6 precursor gas performed on the top of a scanning near-optical field microscope (SNOM) cantilever allows one to create a metal whisker, further used as an atomic force microscopy (AFM) tip to obtain simultaneously SNOM and AFM images of the investigated surface.12 Further, the nanolithography technique could be developed by applying a scanning tunneling microscope attempting to decompose locally the metal carbonyl vapor above a substrate surface by voltage pulses on the tip.13 As a scanning tunneling microscope (STM) allows a tip positioning on the atomic level, it might give atomic-scale information on the mechanism of decomposition and realize a © 2012 American Chemical Society

controlled atom-by-atom cluster growth. For the development of such STM applications for nanolithography by metal carbonyl decomposition, it is important to study the adsorption and film growth of metal carbonyl molecules on appropriate substrates. Metal carbonyl molecules interact through van der Waals forces and form molecular crystals in bulk. It is interesting to study an arrangement of weakly interacting molecules in a twodimensional layer on a noble metal substrate. However, structural studies are difficult to perform using standard techniques, such as X-ray photoelectron diffraction (XPD) and low-energy electron diffraction (LEED), because the carbonyl molecules easily decompose under X-ray and electron beam irradiation, as it was mentioned above. In this paper, we report a detailed study of an adsorbed monolayer structure of molybdenum hexacarbonyl molecules on the surface of Cu(111) by STM and first-principles calculations.

2. EXPERIMENTAL AND COMPUTATIONAL DETAILS The experiment was done in an ultra-high-vacuum (UHV) chamber with a base pressure lower than 1 × 10−10 mbar equipped with a variable-temperature STM (Omicron NanoTechnology). The clean surface of the Cu(111) crystal (Surface Preparation Laboratory, The Netherlands) was prepared by cycles of argon ion-sputtering (0.8 keV, 15 min), followed by annealing at 700 K. The cleaning cycles were repeated until a sharp 1 × 1 LEED pattern of Cu(111) was observed. The welldefined 1 × 1 order and the cleanliness of the surface were confirmed by STM with atomic resolution at room temperature and at 160 K achieved by liquid nitrogen cooling. Received: January 25, 2012 Revised: April 27, 2012 Published: April 30, 2012 10617

dx.doi.org/10.1021/jp300832a | J. Phys. Chem. C 2012, 116, 10617−10622

The Journal of Physical Chemistry C

Article

STM images were taken using etched Pt/Ir tips. The tips were cleaned by ion-sputtering (Ar+, 0.8 keV) and annealed in a vacuum at 600 K. When tip shape effects appeared during the work, interfering with the images, the tip shape was corrected by negative voltage pulses (2−4 V) above the clean Cu surface and/or by mildly touching the surface with the tip. The following conditions were used for the STM images: bias voltage, Vb = 5−300 mV; tunneling current, It = 1−100 pA for the carbonyl monolayer and 1−20 nA for the clean copper surface with atomic resolution. The desorption temperature of Mo(CO)6 molecules on most noble metals, metal oxide, and relatively inert surfaces, such as graphite, is around 200 K.5,8,14 For the Cu(111) substrate, a desorption temperature of 198 K was measured by temperature-programmed desorption.5 Therefore, the copper crystal was cooled to 160 K and exposed to carbonyl vapor on the STM stage. Mo(CO)6 powder was placed into a glass tube and mounted to the STM chamber through a leak valve. The tube was covered by aluminum foil in order to prevent a partial decomposition of the molecules due to light irradiation. The powder was turbo-pumped during 10 min just before exposition in order to reduce the partial pressure of carbon monoxide produced by decomposition. Otherwise, carbon monoxide may poison the substrate, preventing adsorption of the carbonyl molecules directly on copper. The adsorption of the Mo(CO)6 monolayer on Cu(111) is modeled using first-principles calculations based on density functional theory with dispersion force corrections (DFT-D), as proposed by Grimme.15 The plane-wave code VASP is used with projector-augmented wave potentials as provided in the package and the PBE-GGA exchange-correlation functional.16,17 The Kohn−Sham orbitals are expanded in plane waves up to 400 eV. The cutoff radius for the dispersion interactions is set to 15 Å. The Cu(111) substrate is modeled with a slab of four monolayers. The lower two monolayers are frozen in their bulk position, and the upper two are free to relax. Repeated Cu slabs are separated by 21.3 Å of vacuum. Once the Mo(CO)6 molecules are adsorbed, the vacuum space between the molecules and the next Cu slab is still more than 12 Å in all cases. To model single molecule adsorption, we place one Mo(CO)6 molecule in a 4 × 4 supercell of Cu(111), with a lateral size of 10.11 Å. The Brillouin zone is sampled with a Γcentered 3 × 3 ×1 mesh. For the Mo(CO)6 monolayer, both free-standing and adsorbed on Cu(111), we use a √7 × √7 R19 supercell with a lateral size of 6.68 Å and k-points on a 6 × 6 × 2 mesh. Except for the results shown in Figure 3, all atomic coordinates of the molecules are fully optimized.

Figure 1. 330 × 330 Å2 STM image of Mo(CO)6 monolayer on two terraces of Cu(111): tunneling current, It = 8 pA; bias voltage, Vb = 300 mV. Upper inset: 24 × 24 Å2 image of initial Cu(111) surface, It = 3 nA, Vb = 10 mV. Lower inset: 24 × 24 Å2 area of adsorbed carbonyl molecules, It = 12 pA, Vb = 300 mV.

formed on terraces of the cooled Cu(111) surface; see Figure 1. The monolayer has a hexagonal structure with a measured lattice parameter of aMoCO = 6.7 Å (±0.1) and axes rotated by about 20° with respect to those of Cu(111). The observed apparent height of the monolayer is about 1 Å, and the corrugation is 0.4 Å. Both the lattice parameter and the orientation of the monolayer are consistent with a commensurate (√7 × √7)R19 superstructure. This structure is not due to carbon monoxide contamination. CO/Cu(111) system has been quite extensively studied by LEED, and different superstructures have been found depending on the temperature.19 At the temperature of the substrate T = 160 K, CO forms a (√3 × √3)R30 structure, as we have checked in separate experiments. The possibility that the observed monolayer is formed by metallic Mo can also be ruled out because of the low surface mobility of Mo atoms at the experimental temperature and the tendency of Mo to form clusters on noble metal substrates.2 Interestingly, scanning of the adsorbed monolayer under bias voltages and tunneling currents higher than Vb = 100−300 mV and It = 0.1−1 nA (varying for different tips) usually destroys the monolayer ordering. Indeed, an increased bias voltage and tunneling current excites locally the molecules by a moderate electric field and tip−surface interaction. Under such conditions, the scan produces a disordered striped image, indicating an interaction of molecules with the tip. This shows that already moderate values of the bias voltages and tunneling currents give rise to a tip−surface interaction that is larger than forces holding the molecules in the ordered monolayer structure. However, the monolayer destruction by the scanning tip at a moderate bias voltage can be a result of hexacarbonyl molecule decomposition by the tip at applied tunneling conditions. The measured lattice parameter of the monolayer (6.7 Å) is comparable with the molecular distances in bulk Mo(CO)6, which has an approximately closed-packed structure where each molecule has 12 nearest neighbors, 6 of which are at distances of 6.48−6.53 Å, and the other 6 at 7.20−7.26 Å.20 The fact that the nearest-neighbor distance in the monolayer is slightly larger

3. RESULTS AND DISCUSSION The lattice parameter and orientation of the clean copper surface obtained from the STM image (see inset in Figure 1) are used as a reference to determine the monolayer arrangement. The measured values of 2.6 Å for the lateral atom ordering and 2.1 Å for the single step height on the clean surface of the copper crystal are in good agreement with literature values (aCu = 2.56 Å).18 Single carbonyl molecules could not be visualized by STM in the temperature range of the experiment (∼160 K) because of their high mobility under the tip. This indicates weak bonding of the molybdenum carbonyl to the copper surface atoms, which can be easily perturbed by the temperature factor and/or tip−molecule interaction. After 1−2 L of carbonyl exposition, a uniformly ordered monolayer is 10618

dx.doi.org/10.1021/jp300832a | J. Phys. Chem. C 2012, 116, 10617−10622

The Journal of Physical Chemistry C

Article

than that in the bulk could be a consequence of the different orientation of the molecules in the monolayer and/or stronger interaction of hexacarbonyl molecules with the substrate in comparison with the intermolecular forces. To understand the molecular orientation in the monolayer, it is useful to compare the measured monolayer unit cell with the molecular dimensions. The size of the Mo(CO)6 molecule is known from the theoretical and experimental bond length data.21,22 The Mo−O distance is 3.2 Å, which leads to a distance of 4.5 Å between two nearest oxygen atoms in the molecule and 6.4 Å between opposite oxygen atoms; see Figure 2. Three oxygen

Any displacements of the molecule parallel to the surface and rotations around the molecular axis along the surface normal resulted in very small energy variations of 0.01 eV or less. Since the variation is of the same order as the estimated numerical error of the calculations, the exact ground-state adsorption site could not be determined. However, the value 0.01 eV provides a rough estimation of the potential barriers for lateral and rotational diffusion of the molecules on the surface. Such low diffusion barriers correspond to an extremely high mobility of the individual adsorbed molecules, which favors the formation of well-ordered two-dimensional structures at the temperature of the experiment (160 K) and causes the difficulty in observing isolated molecules by STM. The adsorption energy of a single molecule (SM) is defined as Eads(SM) = ESM/sub − ESM − Esub, where ESM/sub, ESM, and Esub are the total energies of the adsorbed single molecule, the free single molecule, and the clean metal substrate, respectively.24 The calculated values of the adsorption energy for the single molecule in the three orientations are given in Table 1. They are in the 0.2−0.5 eV range, which is typical for

Figure 2. Different Mo(CO)6 molecule orientations with respect to the surface of Cu(111). Gray balls, molybdenum atoms; green balls, carbon atoms; red balls, oxygen; yellow balls, copper.

Table 1. Calculated Values of Adsorption Energy of a Single Molecule on the Cu(111), Eads(SM); Cohesive Energy of Free Molybdenum Carbonyl Monolayer, Ecoh; and Energy of Mo(CO)6 Monolayer Formation on the Copper(111) Substrate, Eform

atoms of CO groups from two opposite sides form two parallel planes with a distance of 3.7 Å between them. There are three high symmetry orientations of the molecule on the surface with the surface normal along the 4-fold, 2-fold, or 3-fold symmetry axis of the molecule (Figure 2). In the following, they are denoted as 1-leg, 2-leg, and 3-leg orientations, respectively. We carried out first-principles calculations using the DFT-D approach to understand the molecular adsorption and monolayer growth. To check the performance of this approach with respect to the intermolecular forces, we started by studying bulk Mo(CO)6. The optimized lattice constants (a = 11.93 Å; b = 11.29 Å; c = 6.41 Å) agree within about 1% with experiment (a = 12.02 Å; b = 11.42 Å; c = 6.49 Å).20 The cohesive energy of bulk Mo(CO)6 (defined as the total energy of the bulk minus the energy of free molecules) is Ecoh(bulk) = −0.58 eV/molecule, which compares favorably with the experimental sublimation enthalpy of −0.73 eV/molecule.23 It is worth noticing that DFT without dispersion corrections fails to reproduce the structure and energetics of the bulk Mo(CO)6 crystal. With the pure GGA-PBE functional, the average lattice parameter in the bulk is overestimated by 8% and the cohesive energy (−0.13 eV) is by a factor of 5 too small in comparison with the sublimation enthalpy. With the LDA functional, on the other hand, the cohesive energy (−0.62 eV) agrees with experiment, but the lattice constants are underestimated by 4− 5%. Also for bulk Cu, the DFT-D scheme performs well with a calculated lattice constant of 3.573 Å, which is only about 1% smaller than the experimental value 3.62 Å. The Cu(111) substrate was modeled with a slab of four atomic layers, the upper two of which were relaxed. First, the adsorption of a single Mo(CO)6 molecule on Cu(111) was studied in a 4 × 4 surface supercell (10.11 × 10.11 Å2). We have checked that this cell is large enough that the direct interaction between a molecule and its periodic images becomes negligible. The three principal adsorption orientations shown in Figure 2 have been compared. The optimized distance between the lowest O atom(s) of the molecule and the Cu surface layer is 3.00 (±0.05) Å for all three orientations.

molecule orientation

Eads(SM) (eV)

Ecoh (eV)

Eform (eV)

1-leg 2-leg 3-leg

−0.21 −0.35 −0.44

+0.20 −0.01 −0.31

−0.71

physisorption. The 3-leg orientation of the single molecule is the most stable configuration with the adsorption energy of −0.44 eV, followed by the 2-leg (−0.35 eV) and 1-leg (−0.21 eV) orientations. Insight into the nature of the molecule−substrate interaction can be obtained by a density of electronic states (DOS) analysis. The DOS of the hexacarbonyl molecule on the copper surface is shown in Figure 3, curve a. The difference between this DOS and that of the Cu(111) is shown in Figure 3, curve b. The DOS of free molecule is shown in Figure 3, curve c, for comparison. It can be seen that the molecular orbital structure of the molybdenum hexacarbonyl molecule remains unperturbed after adsorption on the copper surface. This shows that

Figure 3. Density of electronic states (DOS): (a) the molecule on the substrate, (b) the difference of the molecule on the substrate and the clean substrate, and (c) the free molecule. 10619

dx.doi.org/10.1021/jp300832a | J. Phys. Chem. C 2012, 116, 10617−10622

The Journal of Physical Chemistry C

Article

and 2.13−3.56 Å (3-leg). Therefore, for 1- and 2-leg orientations, short O−O distances below 3 Å cannot be avoided, which explains the instability of these structures. For the 3-leg orientation, however, the azimuthal angle can be chosen in such a way that “short” O−O distances are absent. Figure 4a,b shows two arrangements corresponding to φ = 0°

there is no covalent bonding between the molecule and the substrate, confirming the mechanism of physisorption. An important observation is that the highest occupied molecular orbitals (at −2.0 eV) and the lowest unoccupied molecular orbitals (at 1.7 eV) are relatively far from the Fermi level. This fact, together with the instability of the monolayer under scanning at increased bias voltages, makes it impossible to image the molecular orbitals. As a consequence, the orientation of the molecules could not be determined from the STM images. After having discussed the interaction between a single molecule and the substrate, we now turn to the intermolecular interaction within the monolayer. First, we have considered a free-standing hexagonal monolayer of Mo(CO)6 with a lattice constant of 6.684 Å, which is the theoretical value of the (√7 × √7)R19 superstructure on Cu(111). The system was relaxed without symmetry constraints and starting configurations corresponding to 1-, 2-, and 3-leg orientations of the molecules (Figure 2). The structural optimization was stopped when the total energy converged to 10−5 eV. The cohesive energy of the free monolayer per molecule Ecoh characterizing the lateral intermolecular interaction in different molecular orientations is defined as the difference between the energy of the free monolayer EML and the energy of the free single molecule ESM: Ecoh = EML − ESM (see Table 1). A negative cohesive energy means that a formation of the molecular monolayer is favorable with respect to free molecules. As it can be seen from Table 1, only the monolayer with 3-leg oriented molecules is clearly stable with a substantial cohesive energy of −0.31 eV. When the molecules have the 2-leg orientation, the free molecular layer is only marginally stable (Ecoh ≈ 0 eV), and it is unstable with 1leg-oriented molecules. We have thus clearly established that the 3-leg orientation is the energetically most favorable one with respect to both the molecule−substrate interaction and the molecule−molecule interaction within the monolayer. It is interesting to note that, in the 1-leg and 2-leg cases, the molecules show appreciable distortion with some Mo−C−O bond angles deviating from the free molecule case up to 10°. In the case of the free monolayer with the 3-leg orientation, in contrast, the molecules are completely undistorted: the difference of atomic positions between the single molecule and the molecule in the monolayer is below numerical precision (